Recently I decided I would use GIMP for more than quickly changing the colour of something or cropping. It did not go well, if only based on the levels of patience I needed.
To be fair I think a lot of my issues with GIMP does stem from my reliance on the "Photoshop" way. But there are just so many things it also messes up without reason that I would have thought are standard in all image editing software by now.
One of those that baffled me was image boundaries. I have never seen a program enforce a layer boundary which depending on the effect can just straight up crop the image, leaving it all fucky looking.
However I persevered and pretty much got the graphic where I want it. Except that one of the core elements of the graphic is NOISE. Why? Because it used star trails and "galaxies" extensively for the overall design.
This is where another shortcoming of GIMP comes into play, I don't think GIMP is very capable of handling alpha channels very well, not when it comes to feathering, blur, or blend modes. A simple example is how hard the outer section of the red brush is on the image below.
In the below image which is a screenshot you can see the actual view rendering does the fade rather well. So I guess I will need to figure out how to make GIMP export properly, by some miracle.
Aside from that, back to the stars. Now stars are easily created by adding a monochrome noise layer then increasing contrast and if you want some scaling variety you can subsample or start larger and scale down for anything from tiny to large "stars". The contrast makes it so you get a nice spread between the white pixels.
So as you can see I did a few iterations in the above image. with closeup of the worst method to get noise in GIMP for this specific task.
So keep in mind the list for great stars:
Good contrast, spacing.
Not too much clumping.
Size difference.
Hmm, the first two points are similar. Oh well.
Anyway, in GIMP a pretty powerful feature I do appreciate is that almost all effects I add to a layer can be edited after the fact. This is non-destructive editing. So with normal star things you technically destroy a layer to extract white pixels if a blend mode does not assist you with removing the black space.
In GIMP I can remove the black space between white stars in a editable manner.
Ok back to the image below.
Now the image is above.
The default way I found in GIMP to quickly add noise with a immediate star feel is via the "spread" noise effect. This is something I don't recall from photoshop, and as you can see in the image above it just spread the pixels with transparency between.
However the density is too much and to kinda reduce that you would need to spread very far. That is no good because if you are drawing a specific "trail" of stars you want it to likely be about the thickness of your brush.
Then I thought ok, let me duplicate that changing the noise seed and just make it black and have that knock out parts of the white. To an extend it works, but getting rid of the black is a pain in the ass and depending on too many things out of my control I can't really reduce or increase clumping of the white pixels properly.
Then I to play a bit more with the black knockout effect and also applying it to pre-drawn shapes. So I generated a spiral pattern, squished it and boom spiral galaxy?
The main layer effects I used are:
Spread - replace mode
CIE Ich Noise for some more variance on merge mode I think
Threshold to control more of the white pixel density - merge mode
Color to Alpha I think also on merge mode, to remove the black pixels and further control density for translucent pixels.
The color to alpha was really a game changer, since before that I still had some black pixels lurking. Threshold was the other one that really worked well with adding or removing white pixels. Before that I tested levels but considering that I am also using dodge and I have almost all white pixels with varying translucency levels was not able to do shit unless you shift it to the extreme.
Aside from that the only issue I think I can't fix with effects applied directly is the pixel size. I would need to scale these layers up apply and merge the effect in order to scale them down and get variable pixel thickness. Which is anyway the default method I follow but it would have been nice to not have it as a destructive method.
